A recent Goldman Sachs report has raised a rather eye-opening topic: AI is expected to automate 25% of jobs in the United States. At first glance, this percentage doesn't seem particularly alarming, but looking deeper, the situation is far more complex than the surface numbers suggest.

The most surprising point is—white-collar workers are more anxious than blue-collar workers. This challenges our traditional understanding of the automation wave. Over the past 200 years, technological progress mainly targeted physical labor. Jobs like construction and cleaning, although risky, are replaced at a painfully slow pace. The data shows: only 1% of construction cleaning tasks and 4% of maintenance work can be replaced by AI.

However, for knowledge workers, the situation is reversed. Office and administrative support roles have up to 46% of tasks that can be handled by AI, legal professions 44%, engineering and construction 37%. Even business and finance reach 35%, and education and training 27%. This "inverted pyramid" risk structure directly overturns expectations—AI's first large-scale threat is to office workers.

Goldman Sachs predicts that AI will directly displace 6-7% of the US workforce, roughly 9 to 11 million people. They are also quite optimistic, emphasizing that this impact is "temporary," and that AI will create new jobs to fill the gaps.

The problem is—the reality seems to slap their face. By 2025, the average monthly new employment in the US plummeted to 32,000 jobs. What does this mean? Entry-level hiring in the tech industry has dropped by 35%. These data points suggest that AI's displacement effect may be faster and more intense than Goldman Sachs's optimistic expectations.

Let's look at the specific occupational risk rankings. High-risk occupations (more than 30% chance of being replaced) include:

Office and administrative support(46%), legal(44%), engineering and construction(37%), life, physical, and social sciences(36%), business and finance(35%), community and social services(33%), management(32%), sales and related(31%), computer and mathematics(29%), agriculture, fishing, and forestry(28%), protective services(28%), healthcare practitioners and technicians(28%), education, training, and library(27%), healthcare support(26%), arts, design, entertainment, sports, and media(26%).

Medium-low risk occupations (20-25%) include personal care and service(19%).

Truly low-risk occupations (below 10%) are concentrated at the lower end: food preparation and service(12%), transportation and material moving(11%), production(9%), construction and mining(6%), installation, maintenance, and repair(4%), building and grounds cleaning and maintenance(1%).

This contrast is indeed quite bizarre. Traditionally, people worried about physical labor being eliminated by automation, but now it has become a relatively safe domain. Those who should really be worried are high-educated, high-salary knowledge workers—especially those whose tasks are highly standardized and easily processable by AI.